In this paper, we present results from a systematic literature review focused on the status of current scientific knowledge of digital behaviour change interventions for more sustainable food consumption practices. To summarise our findings, it is clear that knowledge about which digital interventions work and why is inconclusive and the field remains to be explored. We carried out a systematic search for peer-reviewed research on the topic, and found a total of 32 papers. Of these, only 15 were included in our corpus (see Figure 1
), since only these had any evaluation of any kind (presenting empirical results) from a behaviour change perspective, even when we set a low requirement for including the studies. It was our conscious choice to stretch the criteria for including papers in this literature review. This decision was taken to ensure that we did not miss any relevant studies. While we have only included studies that present empirical results that make claims in terms of altered food behaviours, all of these 15 studies raise concerns when evaluated from the perspective of whether they led to long-term behaviour change or not. All studies have problems of a methodological nature; the setup of most studies could not even hope to answer the question of if and why they lead to long-term behaviour change, for example due to lack of a control group, lack of a baseline, problems with how variables are defined or measured, length of the study, longitudinal follow-ups, low number of participants, participants not matching the target group or consideration of other possible/likely causes of effects measured. We recognise that the goal of several of these studies was (probably) of an exploratory or “designerly” nature, but we also recognise the need to do follow-up studies examining where actual behaviour change is evaluated in a systematic way.
4.1. Results in Comparison to Other Behavioural Change Studies
Since our interest lies in understanding the current knowledge on digital behavioural change interventions for sustainable food consumption practices, we analysed primary studies in our corpus using the Behaviour Change Wheel (BCW) [10
] framework. We compared the results of this review with three other reviews from different areas using either the Behaviour Change Wheel or the Behaviour Change Technique Taxonomy v1, concerning interventions for sedentary behaviour [14
], transportation behaviour [13
] and gamified mobile apps for health [3
We found that there were primarily two groups of behaviour change techniques that stood out when compared to the other reviews. The first was feedback and monitoring, which was the most commonly used behaviour change technique used in this review, accounting for 25% of the interventions, which can be compared to 16% for sedentary behaviour [14
] and 13% for gamified health apps [3
]. As noted by Edwards et al. [3
], a wide range of studies have shown that feedback and monitoring are effective techniques for achieving behaviour change, and that a clear link can be made to a theoretical base in control theory. Control theory also suggests that goal setting combined with feedback and monitoring is effective for achieving behaviour change [3
]. Therefore, it is interesting to note that of all BCTs coded, the goals and planning category was only coded four times (1%), which can be compared to 28% for the study on sedentary behaviour [14
] and 10% for the study on gamified health apps [3
]. This could indicate that goals and planning is a theory-based type of intervention with high potential that has not yet been explored within the area of digital behaviour change interventions to change sustainable food consumption practices. However, the fact that certain interventions are successful at changing behaviours in one area does not necessarily entail that the corresponding interventions are successful at changing behaviours in a different area. Thus, it might not be valid to compare interventions in sedentary or transportation behaviour to interventions in food-related behaviour. Moreover, food-related behaviour connected to sustainability denotes a broader range of behaviours compared to, for example, sedentary behaviour. Although our review reveals that most of the literature deals with food waste behaviour, it also contains articles relating to significantly different types of behaviour, e.g., food purchasing behaviour.
The mean number of behaviour change techniques used was relatively low (mean = 3.93, median = 3, range 1–9) compared to the interventions in the study on gamified health (median = 14, range 5–22) [3
] and sedentary behaviour (mean = 9.37) [14
]. A previous meta-analysis on internet-based interventions suggested that the number of techniques used was directly related to outcomes [18
]. This indicates that combining various behaviour change techniques (BCTs) in one intervention could be more effective in terms of behaviour change outcomes, but at the same time, it becomes more difficult to draw conclusions about why a particular intervention worked.
4.2. Behaviours Versus Social Practices
It can be argued that digital behaviour change interventions are not enough to achieve reduced environmental impact. Many strategies are based on perceptions that consumer attitudes and actions are rational and that consumers act as individuals [19
]. These perceptions are challenged from a sociological perspective, for example, arguing that instead of using behaviour as a unit of analysis, a social practice lens is more relevant and has greater potential to reduce environmental impact (e.g., [20
]). A focus on individual choices and behaviours can, for example, make us miss the opportunities and constraints that lie in the social structure and context. From the perspective of social practitioners, there are several factors that influence how we as consumers act and perform practices with other people [19
The benefit of using a social practice theory approach is that it is not enough to judge people’s individual choices; we must also see how they are related to many other things on different levels in society [21
]. These may include the design of a store or a city, laws and regulations, and how we value the food. Approaches based on social practice theory attach importance to habits, routines, competencies, social relations, conventions, norms, rules, materials and structures in the environment. It is important to keep the social practice perspective in mind when designing digital interventions within the food area, because it stimulates reflections on their potential and may also emphasise issues of scalability. Furthermore, keeping this perspective in mind may also facilitate an integration of behaviour interventions with structural and systemic changes. Perhaps the outer layer of the BCW, the “policy types” that can be used or that deliver intervention functions, could bridge the gap between individual behaviours and a larger societal focus. However, notably, none of the studies included in our review explicitly touch upon this outer layer of the BCW or reflect on the behaviour change interventions in relation to other levels in society.
4.3. Extending the Behavior Change Technique Taxonomy
One of the papers we analysed (Bandyopadhyay and Dalvi 2017) made use of techniques that were hard to fit into the Behaviour Change Technique Taxonomy v1 [12
]. We use that paper as a starting point for a discussion regarding two proposed additions to the taxonomy as well as some more general reflections about it. The main component of the interactive installation called “The Darker Side”, which Bandyopadhyay and Dalvi (2017) designed, deployed and tested, was a waste bin in a university food canteen that weighed the food that was thrown away and displayed a personalised message consisting of text and an image on a 40 inch screen: “The messages were designed to evoke pain and give a moderate level of aversive feedback. The messages changed from low to high level of aversion depending upon the amount of food thrown (wasted) by an individual.... A minimum threshold of 100 g [helped eliminate] instances of inedible leftovers being [thrown away]” (Bandyopadhyay and Dalvi 2017, p. 239).
This system stood out, since it was the only system that was explicitly aimed at making users uncomfortable by inducing feelings of shame, against the backdrop of “around 3000 children in India dying every day from illness related to poor diet” (Bandyopadhyay and Dalvi 2017, pp. 235–236). When more than 100 g of food was thrown away, the face of a child “staring at the camera” morphed from happy to sad and from healthy to malnourished.
With inspiration from this paper, we suggest that it would be possible in this case to define two more BCTs related to “personal and public shaming” that are not sufficiently covered by the Behaviour Change Technique Taxonomy v1. “Personal shaming” is the shaming of an individual in his/her own eyes by problematising his/her behaviour and mirroring it back (in Bandyopadhyay and Dalvi (2017) by connecting the behaviour of throwing away excessive amounts of food to the plight of starving children). The goal is that the individual will realise, “When I perform this behaviour, I am not a good person”. “Public shaming”, on the other hand, is the shaming of an individual by allowing others to observe the negative feedback the person receives. The exact setup (e.g., how many others could see the screen, etc.) is unclear even from a close reading of that paper, but it is evident that public shaming is very much in line with the intentions of the designers who, for example, applied “Panopticon”, “a feeling of being observed, under surveillance” (Bandyopadhyay and Dalvi 2017, p. 239), as an awareness-raising strategy. The authors also mention that they build upon the results of another paper in our study, the paper by Comber and Thieme (2012), whose design of the “BinCam” system resulted in self-reflection and “feelings of shame”, although that was not their intention.
Public shaming as described above is related to but does not exactly match any of the existing BCTs. We thus suggest that that the group “social support” should be broadened and renamed “social support and disapproval” and that a fourth BCT should be added, “social or public disapproval”. Our proposed definition and an example of this new BCT are as follows:
Arrange or provide emotional social disapproval (e.g., from friends, relatives, colleagues, “buddies”, staff members or strangers) for performance of the behaviour.
Publicly divulge individual travel-related CO2 emissions in the workplace for all employees.
Personal shaming, as described above, is also related to but does not exactly match any of the existing BCTs. Several of the BCTs in the “natural consequences” group could be possible matches (e.g., “information about health consequences”, “salience of consequences” and “information about social and environmental consequences”), but all of these aim for cognitive functions rather than emotional impact. While we considered proposing a new BCT, we instead opted for a minor alteration of an existing BCT, “salience of consequences” (additions in bold):
Use methods specifically designed to emphasise the consequences of performing the behaviour with the aim of making them more memorable or emotionally impactful (goes beyond informing about consequences).
Connect long-haul flights to images of shrinking ice floes and starving polar bears.
It should be noted, that our proposed addition to “salience of consequences” covers the fact that the purpose of the installation “The Darker Side” is not to provide information (as privileged Indians already know that many children are starving in India), but rather to invoke a feeling of shame when individuals gratuitously waste food.
We end this deliberation with a few more general reflections about the Behaviour Change Technique Taxonomy v1 [12
]. One reflection already mentioned above is that existing BCTs typically aim for information-oriented cognitive functions rather than emotional impact. This means that most of the BCTs have active verbs of the type “introduce”, “inform”, “present”, “establish”, “arrange”, “prompt”, “draw attention to”, “raise awareness of”, “observe”, “record”, “monitor”, “provide feedback”, “review”, etc. No BCTs explicitly aim for emotional impact, such as inducing pride, joy, empathy, shame, (emotional) pain or fear.
Another observation is that the origin of the BCTs [12
] might have influenced their form. While the large group of people who had a hand in developing the BCTs came from many different areas, broadly speaking they all had a background in two fields, health and psychology. All existing examples come from the area of health, and that is also where the Behaviour Change Wheel and Behaviour Change Technique Taxonomy have, for the most part, been used to date. Our study of digital behaviour change interventions for sustainable food consumption practices thus belongs to the minority of studies that use these frameworks to study non-health-related behaviour change interventions. This observation opens up interesting developments of the BCTs when applied in non-health-related areas.
Another observation is that only 2 studies were conducted in a non-Western context (India, Indonesia) and of the remaining 13 studies, 2 came from Australia and 11 from various European countries. We also note that we ourselves were challenged by a study (Bandyopadhyay and Dalvi 2017) that was not conducted in a Western, educated, industrialised, rich and democratic, i.e., WEIRD, context [22
]. This study did not shy away from evoking pain and inducing feelings such as shame and guilt as explicit design strategies, or from referring to surveillance and the Panopticon in a positive light. When we attempted to formulate a new BCT, “social or public disapproval”, we were challenged to think of an existing example of a designed behaviour change intervention that attempts to provide emotional social disapproval. Such systems are seldom or never built for a variety of reasons, although some researchers have begun to examine discomfort [23
] or “uncomfortable user experiences”, e.g., to “explore the deliberate engineering of discomfort as a way to create intense, memorable interactions and engage challenging themes” [24
]. Where Benford et al. [24
] work in the intersection of human–computer interactions and performance studies, we imagine that it could be an insurmountable challenge to get ethical approval to conduct medical or health-related studies that explore the effects of “uncomfortable experiences” or that induce shame, pain or fear.